Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study

Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study View ORCID Profile Ângela Jornada Ben , View ORCID Profile Daniel Mwale , Pam Jansen , View ORCID Profile Owen Mtambo , Niek Versteegde , View ORCID Profile Eveline Geubbels , View ORCID Profile Job Calis , View ORCID Profile Jessica Chikwana , IMPALA study team , View ORCID Profile Jobiba Chinkhumba , View ORCID Profile Wendy Janssens doi: https://doi.org/10.1101/2025.10.31.25339214 Ângela Jornada Ben 1 Amsterdam Institute of Global Health and Development , Amsterdam, the Netherlands 2 Department of Health Sciences, Faculty of Science, Vrije Universiteit Amsterdam , Amsterdam, the Netherlands PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ângela Jornada Ben For correspondence: angelajben{at}gmail.com Daniel Mwale 1 Amsterdam Institute of Global Health and Development , Amsterdam, the Netherlands 4 Kamuzu University of Health Sciences, School of Global and Public Health, Department of Health Systems and Policy, Health Economics and Policy Unit , Blantyre, Malawi 5 Training and Research Unit of Excellence (TRUE) , Blantyre, Malawi MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Daniel Mwale Pam Jansen 6 GOAL 3, ’s Hertogenbosch , the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Owen Mtambo 5 Training and Research Unit of Excellence (TRUE) , Blantyre, Malawi PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Owen Mtambo Niek Versteegde 6 GOAL 3, ’s Hertogenbosch , the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eveline Geubbels 6 GOAL 3, ’s Hertogenbosch , the Netherlands PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Eveline Geubbels Job Calis 1 Amsterdam Institute of Global Health and Development , Amsterdam, the Netherlands 7 Department of Pediatric Intensive Care, Emma Children’s Hospital of the Amsterdam University Medical Centers, University of Amsterdam , Amsterdam, Netherlands 8 Department of Pediatrics and Child Health, Queen Elizabeth Central Hospital & Kamuzu University of Health Sciences , Blantyre, Malawi PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Job Calis Jessica Chikwana 8 Department of Pediatrics and Child Health, Queen Elizabeth Central Hospital & Kamuzu University of Health Sciences , Blantyre, Malawi MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jessica Chikwana Jobiba Chinkhumba 4 Kamuzu University of Health Sciences, School of Global and Public Health, Department of Health Systems and Policy, Health Economics and Policy Unit , Blantyre, Malawi PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jobiba Chinkhumba Wendy Janssens 1 Amsterdam Institute of Global Health and Development , Amsterdam, the Netherlands 3 School of Business and Economics, Vrije Universiteit Amsterdam , Amsterdam, the Netherlands PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Wendy Janssens Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract Background Staff shortages, limited training, and inadequate hospital equipment often delay responses to patient deterioration in low-resource settings. The IMPALA continuous monitoring system was developed to support proactive care for critically ill children in such settings. This study evaluated IMPALA cost-effectiveness compared with current practice manual intermittent monitoring from provider and societal perspectives. Methods We conducted an economic evaluation based on a before-and-after cohort of children (0 to 180 months) admitted to Zomba Central Hospital (ZCH) and St. Lukes Hospital (SLH), Malawi (2022 to 2024), where IMPALA was implemented in high-dependency units (HDUs). Targeted maximum likelihood estimation assessed percentage point (pp) differences in mortality, critical illness events (CIEs), disability-adjusted life years (DALYs), and costs (medical, non-medical, indirect). Incremental cost-effectiveness ratios (ICERs) and cost-effectiveness probabilities were calculated for different willingness-to-pay thresholds. Findings At ZCH paediatric ward, 1 840 pre- and 6 255 post-IMPALA children were included; 248 and 736 were admitted to the HDU. Ward mortality decreased (3.7% to 2.8%), with an adjusted 1.9pp reduction (95%CI: - 3.8;-0.6). At ZCH-HDU, mortality slightly increased (8.1% to 9.0%), but IMPALA was associated with an adjusted 9.8pp reduction (95%CI: -26.5;5.0), a 47.1pp decrease in CIEs (95%CI: -52.9;-41.8), and 5.4 DALYs averted (95%CI: -14.2;3.1). At SLH paediatric ward, mortality decreased (4.0% to 2.1%), with an adjusted 1.6pp reduction (95%CI: -3.2;-0.2), a 25.5pp decrease in CIEs (95% CI: -30.1;-20.9), and 1.0 DALYs averted (95% CI: -1.9;-0.1). Provider and societal costs decreased in both wards, but not in the HDU. IMPALA was dominant in wards and slightly more costly but more effective in the HDU (ICERs -$22.5 to $0.4 per life saved). Cost-effectiveness probabilities ranged from 0.8 to 1.0 in wards and 0.3 to 1.0 in the HDU. Interpretation IMPALA was highly cost-effective, reducing mortality by >40%, morbidity by >50%, increasing DALYs averted, shortening hospital stays, and lowering costs, with spillover benefits from HDUs to wards. Funding This project is part of the EDCTP2 programme (grant number RIA2020I-3294 IMPALA) supported by the European Union and Founders Pledge through GOAL3. Evidence before the study Continuous monitoring reduces mortality in high-resource settings, but its effectiveness in low-resource contexts remains uncertain. Systems designed for well-resourced environments may not suit settings with high patient-to-staff ratios, power instability, and limited supplies. In Malawi, qualitative findings suggest that the IMPALA monitoring system – including battery-supported automated digital continuous monitoring devices and local server, a tablet decision support app, and staff training – is feasible and potentially beneficial. However, evidence on the potential impact and cost-effectiveness are lacking. Added value of this study This study provides empirical evidence that the IMPALA monitoring system is a cost- and life-saving alternative to standard care, which relies on manual intermittent monitoring in low-resource settings. The findings indicate benefits, including saving lives, preventing critical illness events, and reducing disease burden, while lowering inpatient and societal costs by shortening hospital stays on paediatric wards as a spillover effect. Implications of all the available evidence This study implies that robust, well-implemented continuous patient monitoring systems can enhance children’s health outcomes and quality of care while reducing costs in a low-resource setting, highlighting the need for its broader implementation to improve paediatric care worldwide. Introduction Globally, child mortality has declined by 59% in the past three decades. 1 Despite this progress, child mortality rates remain high in sub-Saharan Africa (SSA), at 75.8 deaths per 1000 live births in 2019 1 , being 19 times higher compared to high-income countries (HIC), and accounting for 55% of global mortality under 5 years of age. 1 Despite increasing access to healthcare, healthcare facilities in low- and middle-income countries (LMICs) generally lack staff and resources to deliver high-quality care. Up to 15% of overall deaths in LMICs can be attributable to poor quality of care and may be prevented by improving care. 2 An important cause of in-hospital mortality is related to the late detection of deteriorating vital signs. 2 , 3 This is partly due to reliance on intermittent, manual monitoring, which limits healthcare workers’ ability to promptly respond when a child’s clinical condition worsens. 4 – 6 Although patient monitoring systems show benefits in high-income settings, 7 the value and feasibility of their implementation in low-resource contexts is not proven. This is important as concerns exist about the performance of these technologies as their beneficial impact may be hindered by high patient-to-staff ratios, limited trained personnel, unreliable electricity, and scarce consumables and parts. 8 , 9 To address these challenges, a Malawian and European multidisciplinary team co-designed the Innovative Monitoring in PaediAtrics in Low-resource settings: an Aid to save lives (IMPALA) system – hereafter referred to as IMPALA. 10 The IMPALA consortium consisted of 6 academic partners and one social enterprise (GOAL 3). 10 IMPALA is a multicomponent innovation built around a human-centric service model. 11 It includes battery-backed automated digital continuous monitoring devices and local server, a tablet-based clinical decision support application, staff training and ongoing technical support. The app presents patient data using avatars and a traffic light system to support early detection of clinical deterioration and prioritization of care. Tailored for low-resource settings, the system is robust against heat, moisture, insects, and power outages lasting up to 10 hours. It uses reusable sensors to ensure consistent availability and affordability. The tailored implementation strategy includes technical briefings, emergency and critical care training, guidelines on new care processes, all complemented by an online platform with instructional videos. A ward-level “champions” program reinforces system adoption by assigning staff members responsible for supporting training, promoting knowledge transfer, and resolving technical issues, thereby enhancing ownership, sustainability, and quality improvement. 11 Mixed-methods assessments have shown that the IMPALA system is well-accepted by both healthcare workers and caregivers. 12 – 14 Early qualitative evidence from Malawi indicates that the IMPALA system is addressing real barries, 15 but its (cost)-effectiveness was not yet known. This study assesses the impact of the IMPALA system by evaluating its cost-effectiveness compared to standard care, with mortality and occurrence of critical illness events (CIEs) as primary outcomes and disability-adjusted life years (DALYs) as a secondary outcome, using study setting and real-world data. Methods Setting and study population Malawi has an under-five mortality rate of 40 deaths per 1 000 live births. 16 It has a universal healthcare system that provides free access to public care. Zomba Central Hospital (ZCH), one of the country’s four central referral hospitals, serves 830 000 people with approximately 5 000 paediatric admissions annually. 17 The ZCH paediatric ward has 57 beds, including 16 HDU beds, with multiple critically ill children sharing a bed when needed. 15 St. Luke’s Hospital (SLH), is a faith based hospital that serves 90 000 people in the Zomba region and has around 1 000 paediatric admissions annually. 20 Its paediatric ward has 29 beds including 7 HDU beds. Most paediatric admissions occur in the period January-June, when malaria, anaemia and respiratory infections peak. Both hospitals receive government subsidies, and donor support. SLH supplements its funding through patient contributions. The study population included children aged 0 to 180 months admitted to the ZCH paediatric ward, ZHC-HDU, and to the SLH paediatric ward. Study design This is an economic evaluation based on a retrospective before-and-after cohort study, reported according to the Consolidated Health Economic Evaluation Reporting Standards 2022 (CHEERS) statement. 18 Ethical approval was granted by the University of Malawi College of Medicine Research and Ethics Committee (P.01/22/3552, P.02/22/3575 and P.02/24-0607). Data collection At ZCH, clinical and cost data were collected retrospectively from January 1 to June 30, 2022 (pre-IMPALA cohort), and in two periods after the IMPALA implementation: from January 1 to June 30, 2023, and from January 17 to May 31, 2024. At SLH, data were collected retrospectively for both the pre-IMPALA cohort (February 1, 2022 to January 31, 2023), and the post-IMPALA cohort (February 1, 2023, to January 31, 2024). Clinica data were extracted from paper-based and digital medical charts, and ward registers. Health facility cost data were collected through review of expenditure records and interviews with the hospital manager and financial accountant in 2023 (ZHC) and 2024 (SLH), household costs were obtained through surveys with 150 caregivers during their children’s stay in the ZHC-HDU between January 29 and May 31, 2023. In both hospitals, data were entered via tablet by trained research assistants and stored in the Research Electronic Data Capture (REDCap) system. Although IMPALA was implemented in the HDU, data were collected from both the HDU and paediatric ward to capture spillover effects, as the system was expected to influence care processes and outcomes across the entire ward ( Figure 1 ). Download figure Open in new tab Figure 1. The IMPALA strategy to break the vicious cycle of reactive care IMPALA aims to shift from reactive to proactive care through five key mechanisms: (1) Reducing staff burden by automating repetitive tasks, allowing health workers to focus on clinical decision-making. (2) Enabling early recognition of abnormal vital signs through real-time monitoring and alerts. (3) Promoting proactive care by streamlining workflows and providing targeted staff training. (4) Preventing critical illness events, thereby reducing the intensity of care and resources required per patient. (5) Optimising time and resource availability, enabling more efficient care for other patients. Variables collected at admission included age, sex, weight, HIV status, admission date, and diagnosis in both hospitals. Diagnoses were categorized as: respiratory, gastrointestinal, neurological, renal/cardiovascular, systemic or severe infection/inflammation, malaria, malnutrition, hematologic/oncologic, and other ( appendix p25 ). Vital signs (oxygen saturation (SpO2), respiratory rate (RR), heart rate (HR), and temperature) and the Blantyre coma scale (BCS) 21 were collected only for children admitted to the ZCH-HDU ( appendix p29 ). Vital signs were categorized (abnormal/ normal) based on the World Health Organization guidelines. 19 The BCS scoring ranges from 0 to 5 with scores under 5 considered abnormal. 20 Sample size A sample of 1 353 per cohort was needed to detect a 3% difference in mortality and 197 children per cohort were required for a 10% difference in CIEs (α=0·05, 80% power) based on a pilot including 485 children (pre-IMPALA, n=182, and post-IMPALA, n=303) ( appendix p30 ). 21 Standard care Standard care included intermittent manual monitoring of children’s vital signs by nurses four times a day or more often, depending on the child’s condition. 15 Nurses measured SpO2, RR, HR, temperature, and non-invasive blood pressure (NIBP) if needed. 15 In addition, caregivers informed nurses when observing deterioration. 15 IMPALA IMPALA combines battery-supported devices for automated digital continuous monitoring of vital signs (SpO₂, RR, HR, temperature and non-invasive blood pressure measured using reusable sensors), a local server, a tablet app that triggers visual and sound alerts when values reach life-threatening levels, staff training, and ongoing technical support ( Figure 1 ). Staff training was delivered in person, with ward-level champions supporting adoption, troubleshooting, and knowledge sharing. Nine and six IMPALA devices were deployed in HDUs at ZCH and SLH, respectively. The decision to place a child on the monitor was made by nurses and clinicians based on the child’s clinical condition. Effect outcomes Mortality and CIEs were the primary effect outcomes. A CIE was defined as any life-threatening event or life-saving intervention occurring during a hospital stay ( appendix p31 ). DALYs were the secondary outcome as they capture both years of life lost due to premature mortality and years lived with disability resulting from CIEs (detailed explanation in appendix p32 ). Cost outcomes This economic evaluation considered both provider and societal perspectives. The provider perspective included the IMPALA costs and direct medical costs (hereafter referred to as inpatient costs). The societal perspective expanded on this by incorporating household direct non-medical costs (e.g., transportation, food) and indirect costs (i.e., productivity losses due to work absenteeism) related to hospital stays (hereafter referred to as societal costs). IMPALA costs were fixed at US$2·90 per child, estimated as the average system cost across ZCH and SLH. These costs included purchase, implementation, maintenance/repairs, and daily usage, regardless of whether a child was monitored, as the technology would be available even if a child did not need it ( appendix p33 ). Direct medical costs were calculated by multiplying the length of hospital stay in days by the estimated cost per inpatient day, including personnel, medication, food, and other medical equipment. Costs per inpatient day were derived from the ZCH expenditure records and SLH billing ( appendix p34 ). Direct non-medical and indirect costs were estimated based on the ZCH caregiver survey ( appendix p35 ). Direct non-medical costs were calculated by multiplying the length of hospital stay in days by the average daily expenses incurred by caregivers for transportation, food, accommodation, and other out-of-pocket costs (e.g., over-the-counter medicines, soap). Indirect costs were calculated by multiplying the length of hospital stay in days by the average daily absenteeism cost per caregiver estimated from the caregiver survey (i.e., the self-reported amount of income lost due to missed work per one-day hospital stay, appendix p35 ). Facility costs from ZCH were collected in 2023, US dollars ($) and adjusted for Malawi’s 2024 inflation rate. 22 SLH facility costs and direct non-medical and indirect costs were collected in Malawian Kwacha (MWK) converted to $ using the 2024 World Bank PPP rate. 23 Details on cost estimation and unit prices for the cost-effectiveness analysis are in the appendix p36 . Statistical analysis A descriptive analysis of complete and missing data was conducted by study location. Mean imputation was used to handle missing values at hospital admission. 29 Characteristics of children at admission were reported as n (%) for categorical variables and mean with standard deviation (SD) for continuous variables. Given the non-randomized design, cohort differences at admission were tested using t-tests for continuous variables and chi-squared tests for categorical variables. Targeted maximum likelihood estimation (TMLE) was used to estimate the Average Treatment Effect (ATE) between the pre- and post-IMPALA cohorts in percentage points (pp). 24 , 25 The ATE represents the average effect of IMPALA on the entire population of children, regardless of their actual use of the device. It reflects system-wide effects such as improved workflows and earlier recognition of deterioration, making it well-suited for evaluating a multicomponent intervention. TMLE was performed in three steps: (1) fitting g-computation models for mortality, CIEs, and DALYs and costs using relevant covariates to estimate ATEs; (2) fitting a propensity score model to match pre- and post-IMPALA cohorts and calculate inverse probability weights; and (3) updating the ATEs based on propensity scores. Cohort balance was assessed via propensity score density plots. Pre-TMLE comparisons were obtained from linear regression models ( appendix p37 ). Incremental cost-effectiveness ratios (ICERs) were calculated by dividing the ATE in costs between pre-and post-IMPALA cohorts by the ATE in effects. Bootstrapping with 1 000 replications estimated the joint uncertainty around the ATE in costs and effects, with bootstrapped cost-effect pairs plotted on cost-effectiveness planes (CE-planes). 26 For an easier CE-planes interpretation, the ATEs for mortality, CIEs, and DALYs were multiplied by -1 as a positive value indicates improvement in outcomes. Probabilities of IMPALA being cost-effective compared to standard care were estimated using willingness-to-pay (WTP) thresholds of $0, $224, and $448 per unit of effect gained, based on, respectively, 0, 0.5 and 1 time Malawi’s GDP per capita in 2024. 27 Data analysis was conducted using R (R Core Team, 2024, version 4.2.1). Sensitivity analyses Sensitivity analysis used Coarsened Exact Matching (CEM) to match pre- and post-IMPALA cohorts based on admission variables used in the TMLE model ( appendix p37 ). CEM reduces selection bias through exact matching but may reduce power if matches are limited. 28 Also, the Average Treatment Effect on the Treated (ATET) was estimated using seemingly unrelated regressions consistent with covariates used in TMLE. 36 Only children monitored by IMPALA were matched to pre-IMPALA controls when this information was available (ZHC-HDU), as ATET measures the effect among treated individuals rather than the entire population. Role of the funder source The EDCTP2 programme had no role in the design and conduct of the study, data collection, management, analysis, interpretation, manuscript preparation, review, approval, or the decision to submit the manuscript for publication. GOAL3 installed the monitors and trained the staff. Data sharing Anonymized data of this study can be made available for replication purposes. Results Study population At ZCH paediatric ward, 1 840 pre- and 6 255 post-IMPALA children (0–180 months) were included; 248 and 736 (0–73 months), respectively, were admitted to HDU. In the post-IMPALA cohort, 446 children were monitored by IMPALA devices (all in the HDU). Children in the post-IMPALA cohort were older, had more multimorbidity, and a higher prevalence of malaria compared to the pre-IMPALA cohort. At SLH, 930 pre- and 1 126 post-IMPALA children were included (0–180 months). Children in the post-IMPALA cohort were also older and had a higher prevalence of HIV, but lower multimorbidity ( Table 1 ). At SLH, 29·3% of records had missing information on CIEs ( appendix p39–40 ). View this table: View inline View popup Download powerpoint Table 1 Characteristics of pre- and post-IMPALA cohorts by location Effect outcomes At ZCH paediatric ward, observed mortality decreased from 3·7% pre-IMPALA to 2·8% post-IMPALA, with an adjusted 1·9pp reduction (95% CI: -3·8; -0·6). At ZCH-HDU, observed mortality marginally increased from 8·1% to 9·0%, but after adjustment IMPALA was associated with a 9·8pp reduction in mortality (95% CI: -26·5; 5·0), a 47·1pp decrease in the occurrence of CIEs (95% CI: -52·9; -41·8), and 5·4 DALYs averted per admitted child (95% CI: -14·2; 3·1). At SLH paediatric ward, observed mortality decreased from 4·0 to 2·1%, with an adjusted 1·6pp reduction in mortality (95% CI: -3·2; -0·2); a 25·5pp decrease in the occurrence of CIEs (95% CI: -30·1; -20·9), and 1·0 DALYs averted per admitted child (95% CI: -1·9; -0·1) ( Table 2 ). View this table: View inline View popup Download powerpoint Table 2. Effects and cost outcomes by location Length of Stay and Cost outcomes After the IMPALA implementation, the average hospital length of stay decreased in the paediatric wards at ZCH and SLH, leading to reductions in inpatient and societal costs. In contrast, it was slightly higher in ZCH-HDU, leading to increased costs. The primary cost driver was direct non-medical costs (especially travel costs) in both cohorts at all study sites ( Table 2 ). Cost-effectiveness results At ZCH paediatric ward, ICERs were -$6·8 and -$22·5 per life-saved from the provider and societal perspectives, respectively, indicating that IMPALA was more effective and less costly than standard care ( Table 3 ). In the ZCH-HDU, ICERs were $0·3 and $0·4, suggesting that IMPALA was more effective but incurred slightly higher costs per life-saved. In the SLH paediatric ward, ICERs were -$1·2 and -$6·0, showing the same dominance of IMPALA over standard care as in ZCH paediatric ward. The ICERs for CIEs and DALYs followed this same dominance pattern in the paediatric wards and the more-effective-but-more-costly pattern in the HDU. View this table: View inline View popup Download powerpoint Table 3. Cost-effectiveness results by location In the paediatric wards of both hospitals, most bootstrapped cost-effect pairs for all outcomes (65–81%) were located in the southeast (SE) quadrant of the CE-plane ( Table 3 ), indicating that in these settings, IMPALA was generally more effective and less costly than standard care. The probability of IMPALA being cost-effective compared to standard care ranged from 0·8 to 1·0 for all outcomes across WTP thresholds per unit of effect gained, indicating strong evidence that IMPALA is cost-effective at thresholds corresponding to 0, 0·5, and 1 times Malawi’s GDP per capita. In the HDU, most cost-effect pairs for all outcomes (54–64%) were in the northeast (NE) quadrant, suggesting that, among children admitted to the HDU, IMPALA was more effective but also more costly than standard care. The probability of IMPALA being cost-effective compared to standard care ranged from 0·3 to 1·0 for all outcomes across WTP thresholds per unit of effect gained; as WTP increases above zero, the greater health gains with IMPALA are valued more highly than the extra costs, leading to high probabilities of cost-effectiveness. Figures 2 , 3 , and 4 illustrate Table 3 results, including the CE-planes for all outcomes and the three locations individually. All ICERs (red dots) and most bootstrapped cost-effect pairs (blue dots) of the paediatric wards in the two hospitals fell in the SE quadrants; i.e., simulation scenarios indicated a high degree of certainty around results from both provider and societal perspectives. Results of the sensitivity analyses were in the same direction as those of the main analysis ( appendix p41 ). Download figure Open in new tab Figure 2. Cost-effectiveness (CE) planes for mortality from a provider (PP) and societal perspective (SP) at the paediatric ward of Zomba Central Hospital. The CE-planes display the incremental cost-effectiveness ratio (ICER, red dot) and the distribution of 1,000 bootstrapped cost-effect pairs (blue dots). NE = Northeast (IMPALA monitoring system is more costly and more effective than standard care). The change in mortality was multiplied by -1 as a positive value indicates improvement in outcomes. SE = Southeast (IMPALA monitoring system is less costly and more effective than standard care). SW = Southwest (IMPALA monitoring system is less costly and less effective than standard care). NW = Northwest (IMPALA monitoring system is more costly and less effective than standard care). Download figure Open in new tab Figure 3. Cost-effectiveness (CE) planes for different outcomes from a provider (PP) and societal perspective (SP) at the High-Dependency Unit of Zomba Central Hospital. The CE-planes display the incremental cost-effectiveness ratio (ICER, red dot) and the distribution of 1 000 bootstrapped cost-effect pairs (blue dots). The left-hand panels show the results from a PP perspective, the right-hand panels show the SP perspective; the top row shows the mortality CE-planes, middle row critical illness events (CIEs), bottom row. Disability-adjusted life-year (DALY). The changes in mortality, CIEs, and DALYs were multiplied by -1 as positive values indicate improvement in outcomes. NE = Northeast (IMPALA monitoring system is more costly and more effective than standard care). SE = Southeast (IMPALA monitoring system is less costly and more effective than standard care). SW = Southwest (IMPALA monitoring system is less costly and less effective than standard care). NW = Northwest (IMPALA monitoring system is more costly and less effective than standard care). Download figure Open in new tab Figure 4. Cost-effectiveness (CE) planes for different outcomes from a provider (PP) and societal perspective (SP) at the paediatric ward St. Luke’s Hospital. The left-hand panels show the results from a PP perspective, the right-hand panels show the SP perspective; the top row shows the mortality CE-planes, middle row critical illness event (CIE), bottom row Disability-adjusted life-year (DALY). The CE-planes display the incremental cost-effectiveness ratio (ICER, red dot) and the distribution of 1 000 bootstrapped cost-effect pairs (blue dots). The changes in mortality, CIEs, and DALYs were multiplied by -1 as positive values indicate improvement in outcomes. NE = Northeast (IMPALA monitoring system is more costly and more effective than standard care). SE = Southeast (IMPALA monitoring system is less costly and more effective than standard care). SW = Southwest (IMPALA monitoring system is less costly and less effective than standard care). NW = Northwest (IMPALA monitoring system is more costly and less effective than standard care). Discussion Main findings This is the first study to evaluate the cost-effectiveness of a continuous monitoring system tailored to low-resource settings characterized by high patient-to-staff ratios and frequent power outages. IMPALA implementation in paediatric wards was associated with both significantly improved outcomes and cost savings. Mortality rates decreased by 1.6-1.9 pp, a relative reduction of more than 40%. Total costs from both a provider and societal perspective decreased due to shorter hospital stays, and a likely improvement in workflow. In the HDU, the reduction in mortality was not statistically significant, possibly due to limited sample size, but the risk of CIEs fell by more than 50%, corresponding to an estimated 5 DALYs averted per admitted child. HDU costs slightly increased owing to similar lengths of stay pre- and post-IMPALA. Considering the full cohort populations, IMPALA resulted in an estimated 6 286 DALYs averted on 1 164 admissions at ZCH and 2 056 at SLH on 2 056 admissions. Cost-effectiveness analysis showed that IMPALA was consistently dominant in paediatric wards, being both more effective and less costly across all outcomes, with ICERs well below Malawi’s WTP thresholds from both health provider and societal perspectives. This may reflect spillover effects from its implementation in the HDU, possibly because the IMPALA system reduces nurses’ time spent on vital signs monitoring, thus increasing available time to care for other children 12 . In HDUs, new technologies usually raise costs as a trade-off for reducing mortality, preventing CIEs, and averting DALYs. By contrast, IMPALA improved outcomes in HDUs with only a modest increase in costs. Comparison with the literature Our findings expand on existing evidence from high-resource settings. Based on before-and-after studies and one randomized controlled trial in HICs, a recent meta-analysis found that the use of monitoring systems was associated with a 15% relative reduction in mortality risk compared to standard care (pooled RR 0.85, 95% CI 0.72–0.99) and 17% relative reduction in clinical deterioration events (pooled RR 0.83, 95% CI 0.62–1.11). 7 The substantially higher reductions found in our study emphasize the much greater scope for improvement in low-resource settings. IMPALA is a comprehensive system rather than a monitoring device alone, which limits direct comparisons with prior studies. Only a few economic evaluations exist of similar multicomponent monitoring systems, which differ in study population, cost methodology, modelling approach, and WTP thresholds. Nevertheless, our findings are consistent with available evidence. Guinness et al., using a Delphi consensus among experts, estimated the costs per patient of the Essential Emergency and Critical Care approach (EECC, i.e., vital signs intermittent monitoring, oxygen therapy, and intravenous fluids) in Tanzania and Kenya. 29 The EECC cost per patient day ranged from $1 to $33 in Tanzania, and $2 to $37 in Kenya – generally higher than the IMPALA cost of $3 per patient ( appendix p13 ), regardless of length of stay. Our fixed-rate approach helps avoid additional costs for patients requiring extended care. Similar to our results, Shah et al. show a high probability (0.9) of the EECC being cost-effective for critically ill adult patients compared to standard care relative to a WTP threshold of $101 per DALY averted in Tanzania. 30 Limitations This study has some limitations. First, the non-randomized design is prone to bias due to confounders. We addressed observed confounders using TMLE, which provides unbiased estimates even if either the outcome or propensity score model is mis-specified, and the propensity score overlap indicated comparable cohorts. Second, the limited number of IMPALA devices meant that some children lacked monitoring. Although models were adjusted for device access, the effect a full roll-out may be larger. Third, inpatient costs were estimated using a simplified bed-day approach, assuming uniform resource use, which may not capture variability in patient needs; patient-level cost data would be more precise. Fourth, DALYs effect may be underestimated, as disability weights from a limited set of chronic diseases were used and only time in HDU or paediatric wards was considered. Incorporating long-term sequelae through modelling would allow a more comprehensive assessment of the IMPALA’s impact on disease burden. Implications for clinical practice and decision making This study applied advanced statistical methods to data from three diverse settings. We thereby accounted for real-world data complexities and enhanced the generalizability of our results. Findings indicate that IMPALA is consistently dominant in paediatric wards, reducing mortality, CIEs, and DALYs at low cost, with additional benefits in the HDU where CIEs fell by nearly 50% at only modest cost increases. Implementing IMPALA in HDUs of central and district hospitals under real-world conditions appears cost-saving and life-saving for critically ill children, with positive spillover effects on general paediatric wards. Conclusion This observational study showed that implementing the IMPALA continuous monitoring system, tailored to low-resource settings, together with its implementation strategy, was associated with a highly cost-effective impact: a reduction in overall mortality of more than 40%, a reduction in morbidity of over 50%, more DALYs averted, shorter hospital stays, and lower overall costs. Data Availability Anonymized data of this study can be made available for replication purposes. Authors’ Contributions ÂJB was responsible for Methodology, Formal analysis, Investigation, Writing - Original Draft. GR for Writing - Formal analysis, Writing - Review & Editing. DM, OM, PJ, and EG for Investigation, Data curation, Writing - Review & Editing. NV for Investigation, Software, Writing - Review & Editing. JCa for Conceptualization, Software, Project administration, Funding acquisition, Writing - Review & Editing. JCh for Conceptualization, Methodology, Formal analysis, Investigation, Writing - Review & Editing. JeC for Writing - Review & Editing. WJ for Conceptualization, Methodology, Investigation, Resources, Funding acquisition, Writing - Review & Editing, Supervision. All authors read and approved the manuscript. Conflict of Interest The authors declare no competing interests. Acknowledgment We thank Gabriela Riebeek and William Nkhono for excellent research assistance. We thank the IMPALA scientific advisory board (EM Molyneux, MA Ansermino, A Argent, M Hamaluba, M Heys, V Naanyu) for their continuous voluntary support. We further thank the hospital staff, the patients and their guardians for their participation in this study. Appendix 1. List of investigators IMPALA study team Collaborators in the IMPALA study team include: Job Calis MD PhD, Amsterdam Institute for Global Health and Development, Amsterdam, The Netherlands; Christopher Pell PhD, Amsterdam Institute for Global Health and Development, Amsterdam, The Netherlands; Wendy Janssens PhD Prof, Amsterdam Institute for Global Health and Development, The Netherlands; Ângela Jornada Ben MD PhD, Amsterdam Institute for Global Health and Development, The Netherlands; Daniella Brals PhD, Amsterdam Institute for Global Health and Development, The Netherlands; Mark Hoogendoorn PhD Prof, Amsterdam Institute for Global Health and Development, The Netherlands; Michaël Boele van Hensbroek MD PhD Prof, Amsterdam Institute for Global Health and Development,The Netherlands; Job van Woensel MD PhD Prof, Emma Childrens’ Hospital of the Amsterdam University Medical Centers, The Netherlands; Natacha Berbers MSc, Amsterdam Institute for Global Health and Development, The Netherlands; Nina Meels MSc, Amsterdam Institute for Global Health and Development, The Netherlands; Valeria Cristofoli MSc, Amsterdam Institute for Global Health and Development, The Netherlands; Keerthana Raghavan MSc, Amsterdam Institute for Global Health and Development, The Netherlands; Niek Versteegde MD, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Bart Bierling MSc, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Eline Pieck-KleinJan MSc, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Eveline Geubbels PhD, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Lieke de Mare MSc, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Lennart Blom MD PhD, GOAL 3 B.V., ’s-Hertogenbosch, The Netherlands; Michael Levin MD PhD Prof, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Aubrey Cunnington MD PhD, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Myrsini Kaforou PhD; Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Clare Wilson MD, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Diego Estrada-Rivadeneyra PhD, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Shea Hamilton PhD, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Victoria Wright PhD, Section of Paediatric Infectious Disease, Imperial College London, London W2 1PG, UK; Jonathan Sturgeon PhD, Section of Paediatric Infectious Disease, Imperial College London, Norfolk Place, London W2 1PG, UK; Grieves Mang’anda Jr MPH, Kamuzu University of Health Sciences, Malawi; Marrianne Kasiya MPH, Kamuzu University of Health Sciences, Malawi; Arox W. Kamng’ona PhD Prof, Kamuzu University of Health Sciences, School of Life Sciences and Allied Health Professions, Blantyre, Malawi; Daniel Mwale MSc, Kamuzu University of Health Sciences, Malawi; David Chaima PhD, Kamuzu University of Health Sciences, Malawi; Jenala Njirammadzi - Maleta MD, Kamuzu University of Health Sciences, Malawi; Josephine Langton MBChB Ass Prof, Kamuzu University of Health Sciences, Malawi; Jacquline Msefula MSc, Kamuzu University of Health Sciences, Malawi; James Makina MD, Kamuzu University of Health Sciences, Malawi; Jobiba Chinkhumba PhD, Kamuzu University of Health Sciences, Malawi; Lucinda Manda Taylor PhD; Kamuzu University of Health Sciences, Malawi; William Nkhono MSc, Kamuzu University of Health Sciences, Malawi; Margret Havara BSc, Kamuzu University of Health Sciences, Malawi; Alick Vweza PhD, Malawi University of Business and Applied Sciences, Malawi; Brenald Dzonzi BSc, Malawi University of Business and Applied Sciences, Malawi; Chimwemwe Msosa PhD, Malawi University of Business and Applied Sciences, Malawi; Christina Chiziwa BSc, Malawi University of Business and Applied Sciences, Malawi; Lezzie Chirambo BSc, Malawi University of Business and Applied Sciences, Malawi; Theresa Mkandawiri PhD Prof, Malawi University of Business and Applied Sciences, Malawi; María Villalobos-Quesada PhD, National eHealth Living Lab, Public Health and Primary Care Department, Leiden University Medical Center, The Netherlands; Margot Rakers MD, National eHealth Living Lab, Public Health and Primary Care Department, Leiden University Medical Center, The Netherlands; Foteini Klapsaki Bach, National eHealth Living Lab, Public Health and Primary Care Department, Leiden University Medical Center, The Netherlands; Kamija Phiri MD PhD Prof, Training Research Unit of Excellence, Malawi; Alice Likumbo BSc, Training Research Unit of Excellence, Malawi; Jessica Chikwana MD, Department of Paediatrics, Zomba Central Hospital, Malawi; Glory Mzembe MD, Training Research Unit of Excellence, Malawi; Mary Magoya MSc, Training Research Unit of Excellence, Malawi; Timothy Rambiki MD, Training Research Unit of Excellence, Malawi; Martin Mwangi PhD, Training Research Unit of Excellence, Malawi; Owen Mtambo PhD, Training Research Unit of Excellence, Malawi; Christopher Nkhata, Training Research Unit of Excellence, Malawi; Patrick Chalira, Training Research Unit of Excellence, Malawi. 2. Diagnosis definition Diagnoses were collected from paper-based medical records via tablet and stored in the Research Electronic Data Capture (REDCap) system. Diagnoses were categorized into 1=Hematologic/Oncologic, 2=Respiratory, 3=Gastrointestinal, 4=Neurological, 5=Renal/Cardiovascular, 6=Other, 7=Systemic or Severe Infection/Inflammation, 8=Malaria, and 9=Malnutrition as shown in Supplementary Table 1 . View this table: View inline View popup Supplementary Table 1. Diagnosis definition 3. Vital signs Vital signs (oxygen saturation (SpO2), respiratory rate (RR), heart rate (HR), and temperature) and the Blantyre Coma Scale 14 were collected only for children admitted to the Zomba Central Hospital High-Dependency Unit, aged between 28 days and 72 months. Vital signs were categorized as normal or abnormal based on age threshold values determined by the World Health Organization guideline ( https://www.who.int/publications-detail-redirect/978-92-4-154837-3 ). Threshold values are as follows: 1) SpO2: Any measurement below 92% was considered life-threatening or abnormal; 2) RR: for infants ≤ 2 months old, normal rates were between 40 and 60 breaths per minute; for infants between 2 and 11 months old, between 25 to 50 breaths per minute; for children between 1 and 5 years old, between 20 and 40 breaths per minute. Any respiratory rate outside those limits was considered abnormal; 3) HR: for infants between 0 and 1 year old, the normal heart rates were between 100 and 160 beats per minute (bpm); for children between 1 and 3 years old, between 90 and 150 bpm; and for children between 3 and 5 years old, between 80 and 140 bpm. Any heart rate outside those limits was considered abnormal; and 4) Temperature: Any measurement exceeding 38·0 degrees Celsius was considered abnormal. 4. Sample size calculation To calculate the minimum sample size required to detect a difference of 3% in mortality between critically ill children under the IMPALA system compared to usual care, we used the following formula for comparing two proportions: Where: n is the sample size per group. Z α /2 is the Z-score for the desired significance level (i.e., for 95% confidence, Z α /2 = 1 · 96). Z β is the Z-score for the desired power (e.g., for 80% power, Z β = 0.8). p 1 is the mortality rate for the IMPALA monitoring system group. p 2 is the mortality rate for the usual care group. ( p 1 − p 2 ) is the difference in mortality rates (3% or 0·03). R code: # Parameters alpha <- 0·05 # Significance level (for 95% confidence) beta <- 0·2 # Power (for 80% power) p1 <- 0·1 # Mortality rate for the monitoring system group (e.g., 10%) p2 <- 0·07 # Mortality rate for the usual care group (e.g., 7%) delta <- 0·03 # Difference in mortality rates (3%) # Z-scores Z_alpha <- qnorm(1 - alpha / 2) # Z-score for alpha/2 (two-tailed) Z_beta <- qnorm(1 - beta) # Z-score for beta # Sample size calculation n <- ((Z_alpha + Z_beta)^2 * (p1 * (1 - p1) + p2 * (1 - p2))) / delta^2 # Display the result n_per_group <- ceiling(n) # Round up to the next whole number n_per_group We applied the previously mentioned formula to determine the minimum sample size needed to detect a 10% difference in the occurrence of critical illness events (CIEs). # Parameters alpha <- 0·05 # Significance level (95% confidence) beta <- 0·2 # Power (80%) p1 <- 0·1 # Occurrence rate in one group (e.g., 10%) p2 <- 0·2 # Occurrence rate in the other group (e.g., 20%) delta <- abs(p1 - p2) # Difference in proportions (10%) # Z-scores Z_alpha <- qnorm(1 - alpha / 2) # Z-score for two-tailed test Z_beta <- qnorm(1 - beta) # Z-score for power # Sample size calculation n <- ((Z_alpha + Z_beta)^2 * (p1 * (1 - p1) + p2 * (1 - p2))) / delta^2 # Round up to the next whole number n_per_group <- ceiling(n) # Display the result n_per_group 5. Critical Illness Event definition (CIE) A Critical Illness Event (CIE) was defined as the occurrence of any life-threatening event or life-saving intervention during a hospital stay (yes/no). CIE types (i.e., Respiratory, Circulatory, Neurological, Infectious, or Other) and their respective life-threatening conditions were defined as shown in Supplementary Table 2 . View this table: View inline View popup Download powerpoint Supplementary Table 2. Critical Illness Event type and life-threatening condition definitions 6. Disability-adjusted life year calculation Disability-adjusted life years (DALYs) were calculated for each child as the sum of years of life lost due to premature mortality (YLL) and years of life lived with disability (YLD) as shown by Formula 1. The YLL was calculated by subtracting the age of death ( a ) from the life expectancy at birth ( L ) in Malawi (i.e., 62·9 years) . Children who did not die, had zero YLL. The YLD was calculated by multiplying the sum of disability weights ( DW ) related to the CIEs during HDU stay by the duration of the HDU stay in days converted to years ( L 2 ). Disability weights related to CIEs were based on the Global Burden of Disease report as shown in Supplementary Table 3 1 . View this table: View inline View popup Download powerpoint Supplementary Table 3. Critical Illness Event type and disability weights 7. IMPALA system costs 7.1 Zomba Central Hospital Costs calculations for paediatric ward at Zomba Central Hospital for 9 devices and 1 server-unit. Initial Costs: Monitors: $1,500 × 9 = $13,500 Server Unit: $3,000 × 1 = $3,000 Installation Fee: $1,500 Total Initial Costs: $13,500 + $3,000 + $1,500 = $18,000 Service Costs: Monitors: $365 × 9 = $3,285 Server Unit: $365 × 1 = $365 Total Annual Service Costs: $3,285 + $365 = $3,650 Costs Over 7 Years: $3,650 × 7 = $25,550 Costs Over Lifetime: Initial Costs + Service Costs: $18,000 + $25,550 = $43,550 Per Year: $43,550 ÷ 7 = $6,221·43 Per Month: $6,221·43 ÷ 12 = $518·45 Summary Total Cost Over 7 Years for the whole installation: $43,550 Cost Per Year for the whole installation: $6,221·43 Cost Per Month for the whole installation: $518·45 Average children admitted to the paediatric ward yearly: 5,000 Average costs of the IMPALA system per child (2024): $6,221·43/ 5,000 = $1·24 7.2 St. Luke’s Hospital Costs calculations for paediatric ward at St. Luke’s Hospital for 6 devices and 1 server-unit. Initial Costs: Monitors: $1,500 × 6 = $9,000 Server Unit: $3,000 × 1 = $3,000 Installation Fee: $1,500 Total Initial Costs: $9,000 + $3,000 + $1,500 = $13,500 Annual Service Costs: Monitors: $365 × 6 = $2,190 Server Unit: $365 × 1 = $365 Total Annual Service Costs: $2,190 + $365 = $2,555 Service Costs Over 7 Years: $2,555 × 7 = $17,885 Total Costs Over Lifetime: Initial Costs + Service Costs: $13,500 + $17,885 = $31,385 Costs Per Year: $31,385 ÷ 7 = $4,483·57 Costs Per Month: $4,483·57 ÷ 12 = $373·63 Summary Total Cost Over 7 Years for the whole installation: $31,385 Cost Per Year for the whole installation: $4,483·57 Cost Per Month for the whole installation: $373·63 Average children admitted to the paediatric ward yearly: 1,000 Average costs of the IMPALA system per child (2024): $4,483·57/ 1,000 = $4·48 Average costs of the IMPALA system per child (2024) = ($4·48 + $1·24)/ 2 = $2·86 8. Direct medical costs (one-day inpatient costs) 8.1 Zomba Central Hospital The costs of one day at Zomba Central Hospital were estimated based on the annual HDU expenditure at the Zomba Central Hospital in 2023 ( Supplementary Table 4 ) divided by the number of HDU beds (i.e., 16), and subsequently divided by 365 days (i.e., bed-day-based approach). Costs were adjusted to 2024 Malawi inflation using the Consumer Price Index (CPI) ( https://www.nsomalawi.mw/publications/economy/consumer-price-indices-september-2024 ) (i.e., September 2024 = 196·6, February 2023 = 134·7) Bed-day-based cost estimation (2023): ($128,662/16)/ 365 = $22 Bed-day-based cost estimation (2024): $22 x (196·6/ 134·7) = $32 The bed-day-based approach assumes: 1) the care provided per bed-day is consistent across different patients; 2) any fluctuations in staffing levels or resource usage are averaged out over time; and 3) the average occupancy rate is a reasonable proxy for full utilization. The admission-based approach estimated the one-day HDU costs by dividing the annual HDU expenditure at the Zomba Central Hospital in 2023 by the total number of admissions in 2023 (i.e., 4,340). The admission-based approach thus assumes an HDU duration of 1 day, which may be not the case for critically ill children. Admission-based cost estimation (2023): $128,662/4,340 = $30 Admission-based cost estimation (2024): $30 x (196·6/ 134·7) = $44 View this table: View inline View popup Download powerpoint Supplementary Table 4. Cost items used to estimate a one-day HDU stay based on the annual HDU expenditures at the Zomba Central Hospital (2023) 8.2 St. Luke’s Hospital The cost of one day at St. Luke’s Hospital was calculated by dividing each patient’s total hospital bill (in MWK) by their length of stay, converting the result to $ using the most recent available 2024 PPP exchange rate of 1 $ = 382 MWK ( https://data.worldbank.org/ ), and then averaging the per-day cost across all patients. Patient-level cost estimation (2024): $32 9. Zomba Central Hospital Caregivers survey View this table: View inline View popup Download powerpoint Supplementary Table 5. Direct non-medical cost and indirect cost information reported by caregivers 10. Overview: healthcare utilization items and unit prices The table below shows an overview of healthcare utilization and unit prices used in the cost-effectiveness analysis. View this table: View inline View popup Download powerpoint Supplementary Table 6. Healthcare utilization items and unit prices 11. Targeted maximum likelihood estimation (TMLE) Targeted maximum likelihood estimation (TMLE) was used to estimate the Average Treatment Effect (ATE) between exposed and unexposed groups. The ATE represents the average effect of the IMPALA monitoring system on the entire population of children, regardless of their actual use of the device. 1 , 2 TMLE estimates the conditional mean of the outcome (Y) based on intervention (A) and potential confounders (X) while accounting for the probability of A given the observed X 1 , 2 . This method served as the main analysis and followed three main steps: 3 Fitting g-computation models for mortality, CIEs, and DALYs and costs using relevant covariates to estimate ATEs. Fitting a propensity score model to match pre- and post-IMPALA cohorts and calculate inverse probability weights. Updating the ATE estimates based on propensity scores. TMLE provides unbiased estimates even if the G-computation (i.e., the outcome model) or the propensity score model is misspecified, assuming all relevant confounders are captured (i.e., no unmeasured confounding). 1 , 2 G-computation models were fitted, including mortality, CIE, and DALYs as dependent variables, and exposure to IMPALA plus a set of covariates that were significantly different at admission, along with those deemed relevant by researchers. For Zomba Central Hospital’s paediatric ward, ATEs were adjusted for age, sex, HIV status, multimorbidity, malaria, and the interaction between cohort groups and malaria. For Zomba Central Hospital’s HDU, ATEs were adjusted for age, sex, abnormal respiratory rate, abnormal temperature, Blantyre Coma Scale, multimorbidity, monitoring by IMPALA devices, malaria, and the interaction between cohort groups and malaria. The interaction term was included because malaria cases were more frequent in the post-IMPALA period, potentially confounding the association between cohort groups and the outcomes, mortality, and CIE. For St. Luke’s Hospital’s paediatric ward, adjustments were made for age, sex, HIV status, and multimorbidity. The propensity score models were fitted with cohort groups as the dependent variable and a set of covariates that provided the best balance between cohorts. The balance between matched cohorts was assessed using density plots, which compared the distribution of propensity scores. A greater overlap in the distributions indicated more comparable groups. 4 For the Zomba Central’s paediatric ward data, the propensity score model included age, sex, HIV status, multimorbidity, and month of admission. For the Zomba Central’s HDU data, good propensity score distribution overlap was obtained by fitting the propensity score model with age, sex, HIV status, vital signs, Blantyre coma scale, and month of admission as covariates. For the St. Luke’s paediatric ward, the propensity score model included age, sex, month of admission, and multimorbidity. Download figure Open in new tab Supplementary Figure 1. Density plots. Density plots showing relatively good propensity score distributions overlap between pre- and post-IMPALA cohorts in the three study settings. A good overlap means that pre-IMPALA and post-IMPALA data were comparable for the analysis. HDU: High-Dependency Unit. 1 Kreif N, Grieve R, Radice R, Sekhon JS. Regression-adjusted matching and double-robust methods for estimating average treatment effects in health economic evaluation. Health Serv Outcomes Res Method. 2013;13(2):174-202. doi:10.1007/s10742-013-0109-2 2 van der Laan MJ. Targeted Maximum Likelihood Based Causal Inference: Part I. Int J Biostat . 2010;6(2):2. doi:10.2202/1557-4679.1211 3 Luque-Fernandez MA, Schomaker M, Rachet B, Schnitzer ME. Targeted maximum likelihood estimation for a binary treatment: A tutorial. Statistics in Medicine . 2018;37(16):2530-2546. doi:10.1002/sim.7628 4 Markoulidakis A, Taiyari K, Holmans P, et al. A tutorial comparing different covariate balancing methods with an application evaluating the causal effects of substance use treatment programs for adolescents. Health Serv Outcomes Res Method . 2023;23(2):115-148. doi:10.1007/s10742-022-00280-0 12. Missing data descriptives: Zomba Central Hospital – High-Dependency Unit data View this table: View inline View popup Download powerpoint Supplementary Table 7. Missing data 13. Missing data descriptives: St. Luke’s Hospital – Paediatric ward data View this table: View inline View popup Download powerpoint Supplementary Table 8. Missing data 14. Sensitivity analysis View this table: View inline View popup Download powerpoint Supplementary Table 9. Sensitivity analysis by location Footnotes ↵ * shared final authorship REFERENCES ↵ Sharrow D , Hug L , You D , et al. Global, regional, and national trends in under-5 mortality between 1990 and 2019 with scenario-based projections until 2030: a systematic analysis by the UN Inter-agency Group for Child Mortality Estimation . The Lancet Global Health 2022 ; 10 : e195 – 206 . OpenUrl ↵ Kruk ME , Gage AD , Joseph NT , Danaei G , García-Saisó S , Salomon JA . Mortality due to low-quality health systems in the universal health coverage era: a systematic analysis of amenable deaths in 137 countries . The Lancet 2018 ; 392 : 2203 – 12 . OpenUrl ↵ Jacobs B , Hui K , Lo V , Thiede M , Appelt B , Flessa S . Costing for universal health coverage: insight into essential economic data from three provinces in Cambodia . Health Economics Review 2019 ; 9 : 29 . OpenUrl PubMed ↵ Murthy S , Leligdowicz A , Adhikari NKJ . Intensive Care Unit Capacity in Low-Income Countries: A Systematic Review . PLOS ONE 2015 ; 10 : e0116949 . OpenUrl CrossRef PubMed Turner EL , Nielsen KR , Jamal SM , von Saint André-von Arnim A, Musa NL. A Review of Pediatric Critical Care in Resource-Limited Settings: A Look at Past, Present, and Future Directions . Front Pediatr 2016 ; 4 : 5 . OpenUrl PubMed ↵ Slusher TM , Kiragu AW , Day LT , et al. Pediatric Critical Care in Resource-Limited Settings-Overview and Lessons Learned . Front Pediatr 2018 ; 6 : 49 . OpenUrl PubMed ↵ Chong S-L , Goh MSL , Ong GY-K , et al. Do paediatric early warning systems reduce mortality and critical deterioration events among children? A systematic review and meta-analysis. Resuscitation Plus 2022 ; 11 : 100262 . OpenUrl PubMed ↵ Nyandat J , Murthy S , Muttalib F . Challenges and Opportunities for Implementing Pediatric Early Warning Systems in Low- and Middle-Income Countries—Using Resources Wisely . JAMA Network Open 2022 ; 5 : e221553 . OpenUrl ↵ Rakers MM , van Buchem MM , Kucenko S , et al. Availability of Evidence for Predictive Machine Learning Algorithms in Primary Care: A Systematic Review . JAMA Network Open 2024 ; 7 : e2432990 . OpenUrl PubMed ↵ IMPALA study team . IMPALA project. IMPALA project . https://www.projectimpala.org (accessed June 15, 2025 ). ↵ Versteegde N , Calis J. We need to work smarter, not harder - IMPALA system . https://www.medicusmundi.ch/en/advocacy/publications/mms-bulletin/health-workforce-shortage-are-there-potential/kapitel-1/we-need-to-work-smarter,-not-harder (accessed June 15, 2025 ). ↵ Mwale D , Pell C , Chinkhumba J , et al. Implementing the IMPALA continuous monitoring system for paediatric critical care in Malawi: A mixed methods study of barriers and facilitators . PLoS One 2025 ; 20 : e0329265 . OpenUrl PubMed Chirambo LD , Mulaga AN , Thengolose A , Vweza AO , Bierling BL , Calis J . Evaluation of IMPALA 2.0: Addressing Patient Monitoring in Low-Resource Hospitals in Malawi . Med Sci Monit 2024 ; 30 : e945856 . OpenUrl PubMed ↵ Rakers M , Mwale D , de Mare L , et al. Cau tiously optimistic: paediatric critical care nurses’ perspectives on data-driven algorithms in low-resource settings-a human-centred design study in Malawi . BMC Glob Public Health 2024 ; 2 : 80 . OpenUrl PubMed ↵ Mwale D , Manda-Taylor L , Langton J , et al. The role of healthcare providers and caregivers in monitoring critically ill children: a qualitative study in a tertiary hospital, southern Malawi . BMC Health Services Research 2024 ; 24 : 595 . OpenUrl PubMed ↵ UNICEF. Malawi (MWI) - Demographics, Health & Infant Mortality . UNICEF DATA . https://data.unicef.org/country/mwi/ (accessed Aug 21, 2024 ). ↵ Zomba (District, Malawi) - Population Statistics, Charts, Map and Location. https://www.citypopulation.de/en/malawi/admin/southern/MW303zomba/ (accessed March 22, 2024 ). ↵ Husereau D , Drummond M , Augustovski F , et al. Consolidated Health Economic Evaluation Reporting Standards 2022 (CHEERS 2022) statement: Updated reporting guidance for health economic evaluations . Health Policy OPEN 2022 ; 3 : 100063 . OpenUrl PubMed ↵ WHO . Pocket book of hospital care for children: Second edition . 2013 https://www.who.int/publications-detail-redirect/978-92-4-154837-3 (accessed March 12, 2024 ). ↵ Newton CRJC , Chokwe T , Schellenberg JA , et al. Coma scales for children with severe falciparum malaria . Transactions of The Royal Society of Tropical Medicine and Hygiene 1997 ; 91 : 161 – 5 . OpenUrl CrossRef PubMed ↵ Riebeek G , Jornada Ben  , Janssens W. The Effect of an Improved Monitoring System on the Quality of Care at a Pediatric High-Dependency Unit in Malawi . 2023 ; published online June 29. ↵ National Statistics Office of Malawi . https://www.nsomalawi.mw/publications/economy/consumer-price-indices-september-2024 (accessed Feb 16, 2025 ). ↵ World Bank Open Data - GDP per capita Malawi, PPP ($) . World Bank Open Data . https://data.worldbank.org (accessed April 20, 2024 ). ↵ van der Laan MJ . Targeted Maximum Likelihood Based Causal Inference: Part I . Int J Biostat 2010 ; 6 : 2 . OpenUrl PubMed ↵ Kreif N , Grieve R , Radice R , Sekhon JS . Regression-adjusted matching and double-robust methods for estimating average treatment effects in health economic evaluation . Health Serv Outcomes Res Method 2013 ; 13 : 174 – 202 . OpenUrl ↵ Black WC . The CE plane: a graphic representation of cost-effectiveness . Med Decis Making 1990 ; 10 : 212 – 4 . OpenUrl CrossRef PubMed Web of Science ↵ International Monetary Fund, GDP per capita Malawi , 2024 . https://www.imf.org/external/datamapper/NGDPDPC@WEO/OEMDC/ADVEC/WORLD/MWI (accessed Feb 22, 2025 ). ↵ Iacus SM , King G , Porro G . Causal Inference without Balance Checking: Coarsened Exact Matching . Political Analysis 2012 ; 20 : 1 – 24 . OpenUrl CrossRef ↵ Guinness L , Kairu A , Kuwawenaruwa A , et al. Essential emergency and critical care as a health system response to critical illness and the COVID19 pandemic: what does it cost? Cost Effectiveness and Resource Allocation 2023 ; 21 : 15 . OpenUrl ↵ Shah HA , Baker T , Schell CO , et al. Cost Effectiveness of Strategies for Caring for Critically Ill Patients with COVID-19 in Tanzania . PharmacoEconomics Open 2023 ; 7 : 537 – 52 . OpenUrl PubMed View the discussion thread. Back to top Previous Next Posted November 02, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study Ângela Jornada Ben , Daniel Mwale , Pam Jansen , Owen Mtambo , Niek Versteegde , Eveline Geubbels , Job Calis , Jessica Chikwana , IMPALA study team , Jobiba Chinkhumba , Wendy Janssens medRxiv 2025.10.31.25339214; doi: https://doi.org/10.1101/2025.10.31.25339214 Share This Article: Copy Citation Tools Cost-Effectiveness of the IMPALA Monitoring System for Hospitalised Children in Low-Resource Settings: A Pragmatic Before-and-After Study Ângela Jornada Ben , Daniel Mwale , Pam Jansen , Owen Mtambo , Niek Versteegde , Eveline Geubbels , Job Calis , Jessica Chikwana , IMPALA study team , Jobiba Chinkhumba , Wendy Janssens medRxiv 2025.10.31.25339214; doi: https://doi.org/10.1101/2025.10.31.25339214 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Pediatrics Subject Areas All Articles Addiction Medicine (567) Allergy and Immunology (863) Anesthesia (296) Cardiovascular Medicine (4409) Dentistry and Oral Medicine (443) Dermatology (380) Emergency Medicine (606) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1505) Epidemiology (15203) Forensic Medicine (30) Gastroenterology (1119) Genetic and Genomic Medicine (6570) Geriatric Medicine (666) Health Economics (994) Health Informatics (4510) Health Policy (1365) Health Systems and Quality Improvement (1608) Hematology (537) HIV/AIDS (1263) Infectious Diseases (except HIV/AIDS) (15902) Intensive Care and Critical Care Medicine (1103) Medical Education (620) Medical Ethics (144) Nephrology (665) Neurology (6572) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1139) Occupational and Environmental Health (954) Oncology (3319) Ophthalmology (967) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (662) Pediatrics (1689) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5421) Public and Global Health (9205) Radiology and Imaging (2190) Rehabilitation Medicine and Physical Therapy (1367) Respiratory Medicine (1191) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (288) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fe71f139bd01640',t:'MTc3OTIzNTg1OA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();